Membrane Lipids Augment Cell Envelope Stress Signaling via the MadRS System to Defend Against Antimicrobial Peptides and Antibiotics in Enterococcus faecalis

Abstract

Enterococci have evolved resistance mechanisms to protect their cell envelopes against bacteriocins and host cationic antimicrobial peptides (CAMPs) produced in the gastrointestinal environment. Activation of the membrane stress response has also been tied to resistance to the lipopeptide antibiotic daptomycin. However, the actual effectors mediating resistance have not been elucidated. Here, we show that the MadRS (formerly YxdJK) membrane antimicrobial peptide defense system controls a network of genes, including a previously uncharacterized 3-gene operon (madEFG) that protects the Enterococcus faecalis cell envelope from antimicrobial peptides. Constitutive activation of the system confers protection against CAMPs and daptomycin in the absence of a functional LiaFSR system and leads to persistence of cardiac microlesions in vivo. Moreover, changes in the lipid cell membrane environment alter CAMP susceptibility and expression of the MadRS system. Thus, we provide a framework supporting a multilayered envelope defense mechanism for resistance and survival coupled to virulence.

Graphical Abstract

Open in new tabDownload slide

(See the Editorial Commentary by Van Tyne on pages XX–XX.)

Enterococci are a leading cause of health care-associated infections and possess a diverse array of antimicrobial resistance determinants [1, 2]. These organisms are colonizers of the gastrointestinal tract, where they are exposed to cationic antimicrobial peptides (CAMPs) produced by the innate immune system [3]. Several important antibiotics used in clinical practice, including the lipopeptide daptomycin, share functional similarities with CAMPs, and enterococci have evolved several mechanisms to sense and respond to these agents [4–8]. In particular, the LiaFSR system of Enterococcus faecalis has been linked to survival in the presence of daptomycin, the cathelicidin LL-37, and other CAMPs, via a secreted protein (LiaX), which plays a dual sensing and regulatory role mediated by the N- and C-terminal domains, respectively [9]. Experimental evolution in the presence of daptomycin using both E. faecalis and Enterococcus faecium strains lacking the LiaR response regulator and deficient in LiaFSR signaling suggested that activation of the YxdJK 2-component system in the presence of mutations in dak (encoding a putative fatty acid kinase) could also lead to the development of daptomycin resistance [5, 10].

The YxdJK system shares homology with other ABC transporters involved in lantibiotic and antimicrobial peptide resistance, including the BceRS-BceAB system from Bacillus subtilis and the GraRS-VraDE system from Staphylococcus aureus [11]. YxdJK and its downstream effectors function as a transcriptional unit; however, these genes are located in multiple locations throughout the enterococcal chromosome. Dak is a homologue of the fatty acid kinase Fak in S. aureus, responsible for the first step in incorporation of exogenous fatty acids into bacterial phospholipids [12]. Fatty acids have been implicated in increasing survival in the presence of daptomycin and other membrane stressors [13]. As the specific mechanism by which E. faecalis defends the cell envelope against CAMPs is unknown, we sought to determine the role of dak and the YxdJK system (which we hereafter refer to as the MadRS system for membrane antimicrobial peptide defense), in cell envelope protection against daptomycin and host-derived CAMPs.

METHODS

Genetic Manipulation of Strains

For details, please refer to the Supplementary Methods. Bacterial strains and plasmids used in this work are listed in Supplementary Table 1; primers are listed in Supplementary Table 2. Chromosomal deletions were generated using the pCE plasmid and a derivative of OG1RF with the E. faecalis ATCC 4200 CRISPR1 cas9 gene inserted in a neutral genomic insertion site to aid in genetic manipulation, as previously described [14, 15].

RNA Sequence Analysis and qRT-PCR

RNA sequence analysis and quantitative reverse-transcription polymerase chain reaction (qRT-PCR) were performed as previously described [9]. Strains were grown overnight in tryptic soy broth (TSB; Becton Dickinson) plus 50 mg/L of calcium chloride, then diluted into fresh TSB with Ca²⁺ alone or with daptomycin or bacitracin-zinc at 50% of the minimum inhibitory concentration (MIC). RNA was isolated after growth to mid-exponential phase at 37°C. Sequencing was performed on an Illumina HiSeq2000. Reads were mapped using the OG1RF genome. qRT-PCR was performed using SYBR Green in a CFX96 Touch Real-Time PCR Detection System (Bio-Rad; primers are listed in Supplementary Table 2). For all experiments, expression level was quantified relative to the control strain, OG1RF, without antibiotic exposure.

Antimicrobial Peptide Killing Assays

Antimicrobial peptide killing assays for LL-37 (50 µg/mL; catalog No. AS-61302; Anaspec), HBD3 (6 µg/mL; catalog No. AS-60741; Anaspec), and HNP-1 (10 µg/mL; catalog No. AS-60743; Anaspec) were performed as previously described [16, 17]. Assay buffer of RPMI + 5% LB was used for LL-37 and HBD3, and 10 mM KH₂PO₄, pH 7.4 for HNP-1. Mid-exponential phase cultures were diluted in assay buffer to a final inoculum of 1 × 10⁴ colony-forming units (CFU)/mL and incubated for 2 hours at 37°C with or without peptides. Cells were washed and dilutions plated for colony counts. Percent survival for each strain was calculated using the difference between treated strain and growth control.

Determination of Membrane Lipids by Hydrophilic Interaction Chromatography-Ion Mobility-Mass Spectrometry

Lipids were extracted by the method of Bligh and Dyer [18–20]. Lipids were separated on a Waters UPLC using a Phenomenex Kinetex hydrophilic interaction chromatography-ion (HILIC) column (2.1 × 100 mm, 1.7 µm) at a temperature of 40°C and a flow rate of 0.5 mL/minute [21, 22]. Lipid analysis was performed on a Waters Synapt G2-XS platform with calibration of ion mobility (IM) measurements [23]. Data analysis was performed using the Progenesis QI software (Nonlinear Dynamics) for alignment, peak detection, and normalization. Lipid identification was based on m/z (within 10 ppm mass accuracy), retention time, and collision cross-section with an in-house version of LipidPioneer and LiPydomics [22, 24, 25].

Caenorhabditis elegans Infection Model

A C. elegans infection model was used to assay virulence of OG1RF and derivatives in wild-type and pmk-1 deficient nematodes [26, 27]. Synchronized young adult nematodes were infected and 60–90 nematodes were transferred to a liquid media of 80% MW9 buffer and 20% BHI broth in 96-well plates. Plates were incubated at 25°C and evaluated daily for nematode survival. Survival curves from 3 independent runs were plotted using Kaplan-Meier log rank analysis.

Mouse Peritonitis Infection Model

A previously established mouse peritonitis model was used for this study [28]. The protocol was approved by the University of Texas Health Science Center (UTHSC) Animal Welfare Committee (AWC-19-0058). Female 4 to 6-week-old outbred ICR mice (Envigo) weighing approximately 25 grams were inoculated via intraperitoneal injection and followed for 96 hours postinfection. At the time of death, spleen and heart were aseptically excised. Spleens were homogenized, diluted in saline, and plated onto BHI agar for colony counts. To assess for cardiac microlesions, hearts were placed in formalin at the time of removal then processed by the UTHSC pathology core laboratory [29]. Sections were imaged with a Keyence BZ-X700 microscope, the number of microlesions was determined for each section, and mean number of microlesions per section for each strain and the area of each lesion was determined using FIJI software [30].

RESULTS

Identification of the MadRS Regulon

We previously described an alanine to aspartate substitution at amino acid position 202 in the MadS histidine kinase associated with daptomycin resistance in E. faecalis [5]. In the OG1RFmadS_A202E strain, the MIC of daptomycin increased from 1.5 to 6 μg/mL and the MIC of bacitracin from 32 to 96 μg/mL (Table 1). In the OG117 background, introduction of the madS_A202E allele combined with deletion of the dak gene led to an increase in daptomycin and bacitracin MICs. No strains exhibited changes in susceptibility or tolerance to killing with ampicillin or vancomycin, even with physiologic concentrations of bicarbonate (Supplementary Table 3). Because the MadRS system is known to mediate protection from bacitracin via the presence of the ABC transporter MadLM (YxdLM), we hypothesized that this transporter may also be involved in the daptomycin resistance phenotype, similar to VraDE in S. aureus [31, 32]. However, deletion of the genes encoding MadLM from the chromosome of the daptomycin-resistant E. faecalis OG117ΔdakmadS_A202E led to a decrease in the MIC of bacitracin, but not daptomycin (Table 1). Thus, we applied transcriptomics to identify additional genes that could impact daptomycin MIC.

We compared differentially expressed genes in OG1RFΔmadR, which lacks the response regulator of the MadRS system, and OG1RFmadS_A202E to the parent strain OG1RF in the presence and absence of daptomycin (Figure 1A and 1B, and Supplementary Table 2). Comparing OG1RFΔmadR to OG1RF, there were 8 genes with a significant difference in expression without daptomycin exposure. The most strongly downregulated was a previously unidentified putative operon of 3 genes, madEFG, followed by madLM, as well as the genes mprF2 (lysyl-PG-synthetase) and salA (peptidoglycan hydrolase). In OG1RFmadS_A202E, there were 368 and 364 differentially expressed genes with and without daptomycin, respectively. These included increased expression of madEFG, the dlt operon, madLM, salA, and downregulation of upp (uracil phosphoribosyltransferase). We confirmed these results with qRT-PCR for the genes madG, madL, dltA, and madA (previously identified as part of the MadRS network). The madR knockout exhibited decreased gene expression of madG and madL compared to OG1RF, both with and without daptomycin. There was no significant difference in expression of either madA or dltA. The madS_A202E background exhibited a significant increase in the expression of madG, madL, and dltA, particularly on daptomycin exposure (Figure 1C).

The MadRS Network Provides a Targeted Response to Antibiotics and Host-Derived AMPs

To study the effects of each component of the MadRS system, we constructed a series of deletion mutants targeting madLM, madEFG, and dltA in daptomycin-resistant OG117ΔdakmadS_A202E. Deletion of madLM or madEFG had no impact on the daptomycin MIC, while deletion of both led to a modest decrease (12 to 8 μg/mL). In contrast, deletion of dltA resulted in an 8-fold reduction in the daptomycin MIC. Because madEFG did not appear to alter susceptibility to bacitracin or daptomycin but were among the most significantly expressed genes under the control of MadR, we investigated alternative roles for MadEFG.

Given the similarities between daptomycin and CAMPs, we assayed the activity of several representative CAMPs, including LL-37 (cathelicidin), human β-defensin 3 (HBD3), and human neutrophil peptide-1 (HNP1, α-defensin) against the MadRS mutants (Figure 2A and 2B, and Supplementary Figure 1). In the presence of LL-37, OG117ΔdakmadS_A202E exhibited a significant increase in survival versus OG117, and deletion of madEFG, but not madLM or dltA, resulted in increased killing by LL-37. Complementation with madEFG at the native chromosomal location restored bacterial survival. For HBD3, deletion of dak was sufficient to increase survival comparable to the OG117ΔdakmadS_A202E mutant. Restoring the dak gene in the native chromosomal location restored HBD3 killing activity to wild-type levels. Loss of madEFG resulted in a decrease in HBD3 survival by 10.1%, but this was not statistically significant. No change in survival was seen for any strain with the α-defensin HNP1.

MadRS Mediates Survival in the Presence of CAMPs Independent of LiaFSR

We next sought to characterize the distinct roles of MadRS and LiaFSR in the cell envelope stress response. Exogenous N-terminal LiaX has been shown to activate LiaFSR signaling and provide protection against membrane stress even in daptomycin-sensitive E. faecalis isolates [9]. Introduction of the madS_A202E allele into OG1RFΔliaR (lacking activation of LiaFSR due to deletion of the gene encoding the response regulator) was sufficient to increase survival in the presence of LL-37 (Figure 2C), suggesting that a functional LiaFSR response was not necessary for MadRS-mediated protection. Conversely, deletion of the gene encoding the MadR response regulator in OG1RF resulted in decreased survival in the presence of LL-37, an effect that could not be rescued with the addition of exogenous LiaX, despite the strain retaining a functional LiaFSR system. A similar pattern was seen with daptomycin, with decreased survival in the madR deletion strain that could not be rescued by addition of LiaX (Figure 2D). These findings indicated that MadRS ultimately controls the effector systems responsible for CAMP resistance in E. faecalis.

To test this hypothesis, we made targeted deletions of the genes encoding MadEFG and DltA in the OG117ΔliaX background, where the LiaFSR system is constitutively expressed due to loss of the regulatory C-terminal domain of LiaX. Consistent with previous results, the OG117ΔliaX strain showed enhanced survival in the presence of LL-37 as compared to OG117 (Figure 2E). Deletion of madEFG in this background led to a decrease in survival, suggesting this system plays a significant role in membrane defense against LL-37, even in the presence of a constitutively active LiaFSR response. The loss of dltA led to decreased survival after exposure to LL-37 in the OG117ΔliaX background that was not statistically significant. Interestingly, in the OG117ΔliaX background loss of madEFG led to a significant decrease in survival in the presence of HBD3 (Figure 2F), suggesting that the MadRS system can respond to both β-defensins and cathelicidins.

MadEFG did not appear to directly impact daptomycin susceptibility; however, a reduction in the MIC of daptomycin from 12 to 2 μg/mL was observed in the OG117ΔliaXΔdltA strain, and the MIC was restored on complementation of dltA in the native chromosomal location. Because D-alanylation of teichoic acids has been implicated in changes of cell surface charge and daptomycin resistance in S. aureus, we evaluated cell surface charge using a cytochrome c binding assay (Supplementary Figure 2). As compared to OG117, OG117ΔliaX displayed a significant increase in positive cell surface charge. This increase was abolished in the OG117ΔliaXΔdltA strain and restored on complementation of dltA. In the OG117Δdak background, there was an increase in cell surface charge of the dak deletion mutant, but no significant changes across the other strains. Thus, electrostatic repulsion cannot sufficiently explain the phenotypic changes observed in E. faecalis.

Loss of dak Alters the Membrane Lipid Environment and Leads to Activation of MadRS in the Absence of Antibiotic Stress

In the absence of dak, the endogenous type II fatty acid synthesis pathway (FASII) supplies acyl-chains for phospholipid synthesis. Consistent with this hypothesis, OG117Δdak has increased sensitivity to triclosan (4 µg/mL), an inhibitor of the FabI enoyl-ACP reductase of FASII, as compared to wild-type OG117 or the dak complemented strain (both 16 µg/mL). During construction of the OG117Δdak derivatives, we noted that OG117ΔdakmadS_A202E and subsequent strains had a restoration of growth (Supplementary Figure 3). Whole-genome sequencing of these strains showed a serine to tyrosine change at amino acid 36 of FabT, a MarR transcriptional repressor, which binds upstream of the Fab operon (encoding the FASII enzymes). This change would be predicted to fall in the DNA binding domain, potentially impairing the binding of the repressor and resulting in constitutive expression of the Fab operon [33]. Colonies with the mutations in FabT had no change in triclosan susceptibility compared to OG117Δdak (4 µg/mL), consistent with continued reliance on endogenous FA synthesis.

To assess membrane changes associated with the absence of dak, we performed hydrophilic interaction chromatography-mass spectrometry (HILIC-MS) to characterize the phospholipids of OG117, OG117Δdak, the OG117Δdak::dak complemented strain, and OG117ΔdakmadS_A202E (Figure 3A and 3B) during exponential growth and stationary phase. In exponential phase, OG117Δdak had a significant decrease in the abundance of phosphatidylglycerol (PG), lysyl-PG (LPG), and diacyglycerol (DG) as compared to OG117. In both the dak complemented and the OG117ΔdakmadS_A202E strain (with the mutation in fabT), there were no significant changes in phospholipid abundance. In stationary phase, the OG117ΔdakmadS_A202E strain had a significantly higher abundance of PG and LPG as compared to OG117, consistent with the increased expression of mprF2 (the product of which synthesizes LPG) in the madS_A202E mutant. The acyl chain composition of each lipid species in exponential and stationary phase is shown in Figure 3C and Supplementary Table 3. Strains with a deletion of dak showed a relative increase in shorter chain and saturated fatty acids as compared to OG117 during exponential phase growth.

To investigate whether the altered membrane environment associated with the deletion of dak influenced expression of the wild-type MadRS system, we evaluated the expression of madG with and without antibiotic stress in the dak mutant strains during exponential growth (Figure 3D). In the absence of antibiotic stress, there was a 2-fold increase in the expression of madG in the dak deletion mutant as compared to OG1RF, similar to the 2-fold induction of madG expression seen after exposure to daptomycin in all strains. Bacitracin led to a robust expression of madG that did not differ across strains. Taken together, our results suggest that changes in phospholipid composition due to the deletion of dak both directly contribute to CAMP resistance and prime the transcriptional response of the MadRS system.

MadRS Is Associated With CAMP-Dependent Changes in Survival of Caenorhabditis elegans

Using a C. elegans infection model, we examined the role of the MadRS system in protecting enterococci from CAMPs produced by the innate immune system. In wild-type C. elegans, there was an attenuation of the OG1RFΔmadR strain as compared to OG1RF and OG1RFmadS_A202E (Figure 4A and 4B). This phenotype was reversed with complementation of madR on a plasmid with minimal differences between OG1RF and the complemented strain. Next we performed the same survival assay in C. elegans lacking the pmk-1 gene, which encodes an ortholog of the p38 MAP kinase necessary for production of CAMPs. Figure 4C shows that there were no differences in survival between Δpmk-1 worms infected with OG1RF, OG1RFΔmadR, or the complemented strain, supporting the notion that the presence of CAMPs mediates survival of C. elegans infected with E. faecalis lacking MadR, rather than a change in strain virulence.

The madS_A202E Allele Is Associated With an Increased Burden of Cardiac Microlesions in a Mouse Peritonitis Model of Infection

Given the importance of host CAMPs to the innate immune response, we sought to investigate the in vivo impact of the MadRS system in a mouse model of peritonitis. We examined overall survival, bacterial burden in the spleen, and the average number and size of cardiac microlesions as assessed by histopathology of cardiac tissue from infected mice [29]. There was no difference in median survival or bacterial burden in the spleen between OG1RF, OG1RFΔmadR, and OG1RFmadS_A202E (Supplementary Figure 4). However, mice infected with OG1RFmadS_A202E showed an increase in the mean number of cardiac microlesions per histopathological section as compared to OG1RFΔmadR (Figure 4D–H), and a significantly larger lesion area as compared to OG1RF.

DISCUSSION

The potential for cross-resistance between antibiotics and CAMPs of the innate immune system is of particular concern. Vancomycin resistant enterococcal (VRE) colonization of the gastrointestinal (GI) tract has been demonstrated to be a precursor to mucosal translocation and bloodstream infections, particularly in the immunocompromised host [34–36]. In this setting, enterococcal exposure to CAMPs in the GI tract can prime antibiotic resistance and may have implications for the use of daptomycin in treating VRE infections, especially as monotherapy. This study characterizes a defense network that protects the enterococcal cell envelope from CAMPs and peptide-like antibiotics.

Our transcriptional analysis shows that the MadRS system has a much broader cell envelope protective effect than previously known (Figure 5). Consistent with prior results, MadR controls expression of MadLM and also results in an increase in the expression of the dlt operon [37]. In addition, MadR appears to regulate the expression of the peptidoglycan hydrolase SalA, the lysylphosphatidylglycerol synthetase MprF2, and the novel madEFG operon. The madEFG genes share sequence identity with the spr0693-0695 operon in Streptococcus pneumoniae, which encode a MacAB-like efflux pump with LL-37 as a potential substrate [38]. The product of spr0693 was predicted to form a hexameric channel of approximately 155 Å, potentially spanning the peptidoglycan layer. We postulate that MadEFG is likely to function by removing CAMPs from the membrane interface and transporting them past the peptidoglycan layer. Concomitantly, the MadRS regulon alters the cell envelope by D-alanylation of wall teichoic acids via Dlt or production of lysyl-phosphatidylglycerol via MprF to prevent diffusion of CAMPs back to the membrane surface. Interestingly, no obvious homologues of madEFG are present in the genome of Enterococcus faecium DO, suggesting that there are important differences in the mechanism underlying membrane defense between E. faecalis and E. faecium.

Alterations of the membrane bilayer, observed in the dak knockout, may provide a further barrier to effective CAMP action. Loss of Dak function resulted in a membrane lipid fatty acid profile that contained significantly more saturated fatty acids, in part explaining our prior observation of decreased membrane fluidity in the Dak mutant [5]. These changes were sufficient to decrease the activity of HBD3, and also appear to prime the MadRS system by leading to an increase in the baseline transcription of target genes. Thus, the membrane lipid environment has both a direct and indirect role in antimicrobial resistance. These findings may explain why mutations in both phospholipid metabolism and cell envelope stress response networks emerge together in daptomycin-resistant isolates.

The MadRS response appears to complement the global stress response of LiaFSR. Recently, a link between the LiaFSR and MadRS systems was demonstrated in E. faecalis, with activation of the LiaFSR system associated with increased expression of MadRS [39]. MadAB-MadS are likely to function via a flux-sensing mechanism, where the transport activity of MadAB is coupled to the kinase activity of MadS [40]. The data presented here suggest that the emergence of daptomycin resistance via MadRS arises via a decoupling of the MadS kinase and the MadAB flux sensor, as the MadS_A202E mutant showed increased expression of the genes of the MadR regulon, even in the absence of antibiotic stress. This decoupling bypassed the potential transcriptional regulation of the MadRS system by LiaFSR, and activation of MadRS results in a survival advantage against CAMPs, even in the absence of liaR.

The results from the C. elegans and mouse peritonitis models suggest that the MadRS system also plays a major role against CAMPs in vivo, and resistance arising via this pathway may have implications for the clearance of deep-seated infections such as infective endocarditis. Interestingly, we did not see a difference in bacterial burden in the spleen. One possibility is that immune-mediated clearance via macrophages and neutrophils may be preserved via activity of alternate peptides such as HNP-1, the activity of which was not affected in our assay conditions. Because peptides such as β-defensins and cathelicidins are an important part of mucosal barrier defense, this resistance pattern may predispose to invasive infection and bacteremia, particularly in immunocompromised hosts with neutropenia [41, 42]. These studies were performed with a laboratory enterococcal strain, and further work is needed to understand the role of this system in clinical isolates with respect to colonization of the GI tract and subsequent risk for translocation and infection.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The authors thank Rafael Rios for assistance with the RNA sequence analysis and Rachael Whitehead for graphical design.

Data availability. Datasets for the whole-genome sequencing and RNA sequences are available at the National Center for Biotechnology Information (NCBI) under Bioproject Accession number PRJNA1025297.

Financial support. This work was supported by National Institute of Health, National Institute of Allergy and Infectious Diseases (NIH/NIAID) (grant number K08 AI135093 to W. R. M.); NIH/NIAID (grant numbers K24 AI121296, R01 AI148342, R01 AI134637, and P01 AI152999 to C. A. A.); NIH (grant numbers R01 DE027608, R01 AI150045, and R21 AI167124 to D. A. G.); NIH/NIAID training fellowship from the Gulf Coast Consortia, Texas Medical Center Training Program in Antimicrobial Resistance (grant number T32 AI141349 to S. L. E.); NIH/NIAID (grant number R01 A1080714 to Y. S.); and NIH/NIAID (grant number R01 AI136979 to L. X.).

References

Author notes